# Who Will Prompt the Prompts? · 谁来提示提示词

By David Jiang · https://dj.theory-a.com/who-will-prompt-the-prompts · bilingual (English and Chinese), interactive

In Shanghai, a family sells its apartment to pay for an only child's computer science degree abroad. The same year, the models learn to code.

Nobody in that family thinks of themselves as a gambler. Yet they have just pushed every chip they own onto a single hand. Why does a move like that feel like love and duty rather than risk? Why can't working harder rescue it? And what would the alternative look like?

Underneath sits a bigger question. Machines took the grind first, then the code, and now they write prompts for each other. Someone still sits at the top of that stack, deciding what all of it is for. Who will prompt the prompts?

Numbers with a dashed gold underline can be dragged, and every diagram responds to touch.

## You're already all in.

Anyone who has played poker knows what “all in” means: every chip you own, pushed onto one hand. Most people grasp why that's dangerous. Few would put their entire savings into a single stock.

Yet families go all in on their children again and again. A household sells its only apartment to fund one degree abroad: one child, one major, one country, with the payoff a decade away and no way to spread the risk. Some of these bets are forced. Nobody chooses the gaokao; it is the only table in the room. Others are chosen: the school-district apartment bought at the top of the market, the tutoring that eats half a salary, the foreign degree picked because the relatives can read its price tag.

That's why so many people live with a low, constant hum of stress. Whether they like it or not, they are investors, and their portfolio is made of the things they love. Cash is a bet on a currency. An apartment is a bet on a story, as many Chinese families have learned. Even refusing to choose is a choice. There is no square on the board that isn't a bet.

Capital is stored time. Every cathedral was paid for by someone who didn't eat all the grain. A degree, a skill, an audience, your health, the trust people place in you: each is stored time, and each comes with a horizon, an exit or no exit, and a crowd of other people buying the same thing. So before you put money, years or love into anything, ask three questions.

We used to sort work into jobs above the API, which tell software what to do, and jobs below it. AI moves the line up to the prompt: whoever holds the right to ask is safe, and everything below gets commoditized. It moves once more, and that answers the question in the title. Who will prompt the prompts? Whoever owns the capital. Above the capital are the people who decide where stored time goes: the money, the compute, the years. Below it are the people who get allocated. A family that sells its only home for one degree has, without meaning to, volunteered to stand below it.

> Everything you hold is capital. You are not. You are the allocator.

## Working harder loses twice.

When a bet like that starts to go wrong, the instinct is to double down. The graduate who can't find a job enrolls in a second master's. The parent with a little savings left opens a bubble tea shop, because bubble tea is hot. Both moves lose, for two different reasons.

In finance, every investment has to beat the risk-free rate: the return you could earn by doing nothing, safely, such as lending to the government. Something new has happened to that rate. a16z estimated that the price of a given level of model capability has been falling about tenfold a year. Money sitting still is getting smarter: $100 spent on thinking from now buys roughly what buys today. Anything built from cognitive work, a skill, a degree, a piece of software, now has to outrun that curve.

Spend $100,000 this year hand-building what next year's model does for $100, and you weren't investing. You were racing the labs with your own money. Even the people funding AI see it: Yishan Wong, Reddit's former CEO, wrote that "every AI application startup is likely to be crushed by rapid expansion of the foundational model providers."

This is what your second master's is competing with. The smiley face costs $20 a month. The thing behind it gets cheaper every quarter and never sleeps.

Picture a street with one successful bubble tea shop. Ten people write business plans for the same street, and every plan looks great, because each is modeled on that first shop. They all get loans and they all open at once, and then the part none of the plans modeled arrives: each other. Ten shops bid for the same storefronts, suppliers and part-time workers, and split the same customers ten ways. Build the street yourself:

Past the dashed line, an owner can work fourteen-hour days, seven days a week, and still be destroying money. Every hour returns less than the capital tied up in the shop costs. Scott Alexander called this kind of race to the bottom Moloch: the god who demands that everyone sacrifice a little more just to stay in place. In Chinese, Moloch has a shorter name. 内卷.

> Idle money is getting smarter. Crowded effort is getting cheaper.

## Bitterness is a cost, not a currency.

Here's the strange part. The owner can read those numbers. She keeps the shop open anyway, and her relatives respect her more for it. In the culture she grew up in, pain is how you measure worth. A shop that loses money fourteen hours a day isn't a bad investment. It's proof of virtue.

And the twist is that eating bitterness really did work, for a long time. A century of farmers, a generation of gaokao students, parents who built a middle class out of nothing: their bitterness paid. But not because pain is valuable. It paid because the return on their effort was positive. In an economy short of skills and roads and factories, every extra hour built something. The bitterness was the cost. The return came from the investment.

Children used to be part of that return. On a farm, a child was an extra pair of hands by seven and the family's pension by forty. Family needs and economic needs pointed the same way: more children meant more labor and more security in old age. In a city apartment the arithmetic flips. A child is mostly cost, for decades, and the return a family can see is status: the school, the score, the degree, a job title you can say out loud at a banquet. The instinct to put everything into children survived the move to the city. The return that justified it didn't.

When the return turns negative, the cost stays. Goodhart's law says that when a measure becomes a target, it stops being a good measure. When pain becomes the measure of worth, it stops measuring worth. And the myth has a second line of defense: we only hear from the planes that came back.

None of this makes the parents fools. They were the best investors of their era: they bought hard work at the bottom, when it was cheap, and it paid off enormously. Chesterton's fence says don't tear a fence down until you know why it was built. This one was built against scarcity. Knowing why it was built is also how you notice when the scarcity has moved.

> When pain becomes the measure of worth, it stops measuring worth.

## Wu wei: getting paid not to do the wrong thing.

The culture under the most pressure to act also wrote the best words against it. Govern a great country the way you fry a small fish: keep flipping it and it falls apart. Who can make muddy water clear? Let it be still. Charlie Munger said it in an American accent: the big money is not in the buying and selling, but in the waiting.

Who can make muddy water clear? Let it be still. Stop scrolling and take your hand off the mouse or the screen. Anything you do stirs it up again.

Wu wei doesn't mean doing nothing. It means not forcing, and in markets, not forcing gets paid. Gold is the purest case. It yields nothing, builds nothing and does nothing, and over long stretches it has held its value while the things people rushed into didn't. Gold pays you for exactly one thing: not doing the wrong thing. In the economist Albert Hirschman's terms, it's exit rather than voice. Instead of staying in a failing game and pushing harder to change it, you quietly step out, and the asset pays you for leaving. China's Gen Z seems to have rediscovered this on its own, collecting one-gram gold beans in little jars.

Somewhere along the way, the culture that wrote the Dao De Jing came to treat stillness as a moral failure. Call it the striver curse. It begins as good sense. In a poor family, rank is safety, and the exam is the one door the family can push on. So the child learns a few rules early: if I rank well, I am real. If I fall behind, I endanger the family story. If I rest, someone else is turning my rest into their advantage.

The curse is that the rules outlive the poverty. Every gate opens onto a waiting room for the next gate: the good school, then the good job, then the apartment, then the second apartment. Winning never turns into safety, only into a ticket for the next tournament. Rest feels faintly disgusting, like falling behind in real time. The striver isn't chasing success. He can't stop, and he has been taught to call that inability a virtue.

It isn't only China. Singapore runs another version of the same machine, and the Singaporean writer Hoeflator has described it from the inside. Children “fight in the books with other 12 year olds for your PSLE examinations,” the exam that sorts them at twelve. Adults are “far too busy trying to make your iron rice bowl to paint, write poetry, or spend hours on an instrument.” His verdict: “In scarcity, the mind cannot flourish, and the brain cannot bloom.” The cycle closes on the next generation, as the middle class buys condo after condo, “banking on the next generation in line to be the exit liquidity.” Korea runs the same machine with cram schools, Silicon Valley with performance reviews. The rules are local. The waiting room is the same everywhere.

Then there's the counterfeit. From the middle of the curve, 躺平, lying flat, looks just like wu wei: both do less. But lying flat walks away from the table. No bets, no allocation, a quiet refusal, and your options expire worthless. Wu wei stays at the table and stops forcing: small bets, capped downside, open upside, held long enough to matter.

> Lying flat sells your options. Wu wei holds them.

## The boat was already there.

There's an old joke about a man caught in a flood. He climbs onto his roof and prays for God to save him. A rowboat comes by, and he waves it away: God will save me. A motorboat comes, and he waves that away too. A helicopter lowers a ladder, and he refuses. He drowns. In heaven he asks God why He didn't help. God says, “I sent you two boats and a helicopter. What more did you want?”

The boats are already here. Anyone with a brokerage account can own a slice of the systems that are squeezing them. Instead of opening the eleventh tea shop, you can own shares in the company that sells syrup and cups to all ten. Mixue, the chain behind tens of thousands of franchised shops, listed in Hong Kong in 2025, and almost all of its revenue comes from selling goods and equipment to its franchisees. On the boba street some owners win and some lose; the supplier gets paid either way. At the scale of a nation, the Three Gorges Dam is run by a listed company, China Yangtze Power. For a few thousand yuan, a family can own a sliver of the largest power station on earth and collect its dividends.

Not investment advice. Companies and funds are named only to illustrate an idea, not as recommendations. Any investment can lose money, and past returns don't predict future ones.

Why wave the boats away? Partly because they don't look like rescue: owning things doesn't hurt enough to feel like work. And partly because we're waiting for a different miracle: the next new thing, the startup that will “put a dent in the universe,” the one big break. Seeing a boat for what it is takes quieter virtues: attention to what is actually in front of you, gratitude for it, and the humility to accept a slow, boring rescue.

Deadlines make it worse. When you need a certain amount of money by a certain age, enough for the apartment before the wedding, or to have “made it” by thirty-five, a steady 20% a year looks like failure. It would turn 100,000 into more than 600,000 in ten years, but not by the deadline. So people reach for the Hail Mary: leverage, meme coins, the all-in. These bets often lose money on average. They only make sense if missing the deadline feels like death. When a life is organized around milestones a few years apart, volatility starts to look rational even when its expected return is negative.

Chinese investors already have a word for what usually happens next: 韭菜, chives. Retail investors are the crop. Every boom grows them, every bust cuts them down, and they grow back for the next harvest. The cynical conclusion is that the market is a casino run for insiders. But look at who gets cut. It is rarely the patient owner of something boring, holding a slice of the dam for ten years. It is the person who chased last month's winner, bought on a tip, borrowed to do it, and sold in a panic. Chives get harvested because they all grow to the same height at the same time. It's a crowded trade again. The market harvests behavior, not ownership.

From the outside, the chive and the patient owner look identical, and so do investing and gambling. Both buy assets, take risk and stare at prices. What separates them is what wuxia novels call 心法, the inner method behind the moves: how much to bet, how long to hold, and the first rule of every martial art, which is don't die. Practicing the moves without the method has a name too. 走火入魔: the fighter torn apart by his own power.

> The boat is already here. Get in, slowly.

## Eat people, or save the children.

Run the opening bet again, the other way. The family keeps the apartment as its floor. Part of the savings goes into things that keep working while everyone sleeps: a slice of the dam, a slice of the syrup company, a jar of gold beans. The rest goes on a few small bets on what the child is actually drawn to, cheap to try and cheap to drop. When computer science turns out not to fit, nobody calls the child lazy. A bad fit is market risk, not a moral failing.

In 1918, Lu Xun published “A Madman's Diary.” Its narrator opens the history books and finds every page covered with the words “benevolence, righteousness and morality.” Unable to sleep, he reads between the lines all night and finds two words written everywhere: 吃人, eat people. A system that eats people doesn't look cruel from the inside. It looks like love, duty and sacrifice. The child is fed, educated and loved, and also turned into the family's pension, its status, and its proof that the suffering meant something.

A year later, in an essay on how fathers should raise their children, Lu Xun described the way out: a parent who carries the burden of the past, “shoulders the gate of darkness,” and lets the children through “to a wide and bright place.” Everything you hold is capital. You are not, and neither is your child. You can invest in a child. You can't own one.

Breaking the cycle is the point of all of this. The striver curse passes down because each generation pays its children in pain and calls it love. Involution reproduces because everyone is too exhausted to step out of the crowded trade. When a child counts as a failure unless she escapes ordinary life, fewer people want to enroll one in the tournament at all. The machines are finally offering a way out of the grind. The hard part is accepting it, and letting our children accept it too.

“A Madman's Diary” ends on a question and a plea. “Perhaps there are still children who have not eaten people? Save the children.”

> Save the children.